Professional Documents
Culture Documents
PII: S1359-1789(20)30224-X
DOI: https://doi.org/10.1016/j.avb.2020.101520
Reference: AVB 101520
Please cite this article as: H.D. Butler, J.W. Caudill, J.M. Craig, et al., 99 Percenters: An
examination of the misconduct careers of the Most violent and disruptive incarcerated
delinquents, Aggression and Violent Behavior (2020), https://doi.org/10.1016/
j.avb.2020.101520
This is a PDF file of an article that has undergone enhancements after acceptance, such
as the addition of a cover page and metadata, and formatting for readability, but it is
not yet the definitive version of record. This version will undergo additional copyediting,
typesetting and review before it is published in its final form, but we are providing this
version to give early visibility of the article. Please note that, during the production
process, errors may be discovered which could affect the content, and all legal disclaimers
that apply to the journal pertain.
99 Percenters: An Examination of the Misconduct Careers of the Most Violent and Disruptive
Incarcerated Delinquents
H. Daniel Butler
Jonathan W. Caudill
of
ro
Jessica M. Craig
Chad R. Trulson
Abstract
Jo
The function of correctional institutions is to maintain institutional security and safety, and
institutional misconduct undermines that security. Few studies have examined the individuals
who pose the greatest threat to institutional security and are responsible for the greatest number
of misconduct incidents (i.e., the 99th percentile of misconducts). This study examines youthful
offenders incarcerated in Texas state juvenile facilities who engage in the most frequent and
most violent misconduct incidents through a series of logistic regression and ROC-AUC models.
The findings indicate that younger age, African-American status, gang activity during
Journal Pre-proof
confinement, being a danger to self, and involvement in multiple forms of treatment evince a
higher risk of engaging in the 99th percentile of total and violent misconduct regardless of model
specification. Other factors such as commitment offense, adverse childhood experiences, and sex
have differential effects depending on whether the 90th or 99th percentile thresholds are
considered. Youth in the 90th percentile account for 49% of total misconduct and 50% of violent
misconduct. Youth in the 99th percentile account for 11% of total misconduct and 51% of violent
misconduct. We discuss implications for policy and practice and the importance of
of
understanding antisocial development of youthful offenders and their institutional
ro
maladjustment.
-p
Keywords: prison violence, institutional misconduct, prison, youth, criminal career, career
re
criminal
Introduction
lP
In February 2020, Jonathan Watson, a convicted murderer serving a life sentence in the
na
indicated that he was experiencing acute homicidal ideation toward other inmates in his
ur
dormitory, and in particular, inmates that were convicted sexual offenders. After experiencing
Jo
distress over the homicidal impulses, Watson requested a transfer to a more restrictive
classification setting. His request was denied in part because correctional counselors doubted the
authenticity of Watson’s violent intentions. The next day, Watson fatally bludgeoned two
inmates because he was outraged about their conviction offense: aggravated sexual assault of a
child. After killing the other prisoners, Watson promptly confessed to correctional officers and
has since provided a written confession. The case awaits formal adjudication and there is no
Although this seems a sensationalistic case, homicide and other forms of serious misconduct are
relatively prevalent in correctional facilities. Based on data from the Bureau of Justice Statistics
Mortality in Correctional Institutions report (Carson & Cowhig, 2020), nearly 1,200 inmate
murders and more than 3,500 inmate suicides occurred in state and federal institutions since
2001. Moreover, these data do not include the thousands of incidents of other serious misconduct
including rioting, assaults, sexual assaults, arson, threats toward correctional officers and
of
Pathological Adjustment to Prison
ro
For decades, penologists studied inmates who had difficulty adjusting to prison life and who
-p
engaged in diverse forms of violent and antisocial conduct that threatened the security of prison
re
institutions (e.g., Adams, 1977; Coe, 1961; Petersilia, Honig, & Hubay, 1980; Poole & Regoli,
1979; Toch, Adams, & Greene, 1987; Wolfgang, 1961; Wood et al., 1966). Although researchers
lP
employed a variety of analytical techniques and data sources, two consistent themes emerged
na
from their research. First, there was a general association between lack of adjustment in society,
as evidenced primarily by criminal history, and subsequent lack of adjustment and involvement
ur
in misconduct in custody. In sum, the most problematic free citizens tended to be the most
Jo
problematic prisoners—this later became known as the importation model of inmate behavior
(Irwin & Cressey, 1962). Second, there was recurrent evidence of a subset of inmates who were
extremely difficult to manage and posed significant threats to other inmates and staff (Marquart
& Crouch, 1984; Wood et al., 1966; Zink, 1957). For instance, in a seminal investigation, Fox
1
In terms of correctional policy, nearly all state departments of correction and the federal Bureau of Prisons built
prisons (e.g., Alcatraz, USP Marion, USP Florence ADMAX) designed specifically to house the most challenging
inmates. Although there is variation across jurisdictions (see, Butler, Griffin, & Johnson, 2013; Butler, Johnson, &
Griffin, 2014), the most consistent criteria for placement in administrative segregation in super-maximum prison
facilities include inmates who have murdered staff, murdered other inmates, sexually assaulted staff, sexually
assaulted other inmates, are gang leaders, have extensive self-injurious behavioral history, and who otherwise have
extensive misconduct careers (Butler et al., 2013; Butler et al., 2014; Lanes, 2011; Lovell, 2008).
Journal Pre-proof
(1958) observed that institutional misconduct is the most pressing problem in correctional
settings and there is great variation in adjustment to prison life as well as willingness to abide by
the individual offender who builds up a series of misconduct reports within the
of
much the same manner in which a chronic neurosis seems to develop.
ro
That quotation attests to the role of self-regulation, psychopathy, and general psychopathology as
-p
important variables to understand misconduct among the most severe prisoners.
re
In the ensuing decades, others have also examined the most difficult-to-manage inmates. For
prisoners with good adjustment using data from Ohio inmates. They found several factors that
na
were associated with chronically disruptive conduct including more extensive criminal careers
(e.g., early onset of conduct problems, multiple juvenile police contacts, prior confinement and
ur
psychopathy (e.g., use of aliases). Based on data from female prisoners, Thomson, Towl, and
Centifanti (2016) reported that women convicted of serious violent crimes, such as murder and
who had pronounced psychopathic features, particularly callousness and impulsivity, were most
likely to constitute threats to the correctional population with their recurrent misconduct and
More contemporary research employed latent class analyses to identify subgroups of inmates by
misconduct risk (e.g., Cihan, Davidson, & Sorensen, 2017; Cihan & Sorensen, 2019; Morris et
Journal Pre-proof
al., 2012), and although disparate data sources have been used, there is once again recurrent
evidence of subgroups of inmates who pose significant problems for correctional officials. To
illustrate, Reidy, Cihan, and Sorensen (2017) studied 915 female Arizona prisoners and found
two subgroups of difficult inmates. The first constituted 8% of the sample and exhibited early-
onset misconduct careers with the highest levels of serious infractions that quickly declined
throughout their incarceration. A smaller group constituted 6.2% of the sample and had the
opposite trajectory: they had delayed-onset of serious misconduct that was consistent with
of
escalation in behavior problems. Several factors were associated with these severe subgroups
ro
including younger age at commitment, prior incarcerations, more mental health problems, greater
-p
violence risk, and more restrictive custody level. In their study of more than 8,000 inmates
re
selected from a northwestern state prison system, Cihan, Sorensen, and Chism (2017) reported
that 8.3% of inmates were “escalators” that started their confinement with compliant conduct that
lP
gradually built to more serious and frequent misconduct violations. Higher rate of prior arrests,
na
having a sexual assault commitment offense, and racial/ethnic minority status were predictive of
Despite a growing literature on the institutional misconduct behavior of adult offenders, less is
Jo
known about intractable conduct among institutionalized delinquent offenders. This gap is
surprising since younger age is one of the most consistent predictors of maladjustment to prison
(Kuanliang, Sorensen, & Cunningham, 2008; Valentine, Mears, & Bales, 2015). Similar to
research using adult samples, prior research indicates that among juveniles there is also evidence
of a subset of severely maladaptive and violent wards; however, their risk profile is less clear.
Drawing on data from 2,520 serious and violent state institutionalized delinquents, Trulson,
DeLisi, Caudill, Belshaw, and Marquart (2010) found that a small cadre of delinquent wards
Journal Pre-proof
produced the majority of violent incidents and misconduct. Specifically, 9.2% of wards
accounted for 21 or more major violations whereas nearly 60% of wards accumulated between
zero and four major violations. These findings suggest that institutional misconduct was highly
skewed and that fewer than one in ten inmates accounted for the bulk of violations among
incarcerated state delinquents. Moreover, Trulson and colleagues found that delinquent history
variables were consistently predictive of major violations (also see, DeLisi et al., 2011). Drawing
on data from 390 juveniles selected from the Oregon Youth Authority, Reidy, Sorensen, and
of
Cihan (2018) found that 20% of juvenile wards were very problematic while in custody and had
ro
early and sustained infractions. Conviction for violent offenses including murder, robbery, and
-p
assault, earlier age of commitment, and Hispanic ethnicity were predictive of this problematic
re
subgroup.
Current Focus
lP
Although penological history is replete with examples of violent and disruptive inmates, a basic
na
criminological profile of the most challenging adult inmates lacks consensus. We know even less
with respect to the institutional misconduct careers of the most prolific state institutionalized
ur
delinquents. The current study sought to fill this void by examining the total and violent
Jo
delinquents. In particular, this study focuses on the misconduct careers of the 99th percentile of
offenders in our sample (with sensitivity models using the 90th percentile).
This focus on the most chronic and violent misconduct perpetrators within a group of
already serious delinquent offenders is important on a number of levels. At the practical level,
associated with removal or severe restriction from beneficial institutional programming and
Journal Pre-proof
treatment—treatment and programming that are perhaps especially important with respect to the
deep-end offenders under study. Understanding the factors related to misconduct perpetration is
important then in efforts to help redirect offenders to more prosocial ways of managing the
institutional regime, to glean as much as they can from the institutional experience, and
ultimately, to help promote positive adjustment to free society once released. Indeed, previous
research has found that institutional misconduct influences post-release recidivism (Cochran &
Mears, 2017; Cochran, Mears, Bales, & Stewart, 2017; Trulson et al., 2010; Trulson, DeLisi, &
of
Marquart, 2011), and additional knowledge on the factors relevant in explaining this behavior
ro
may shed additional light on ways to improve the release trajectories of such serious juvenile
offenders.
-p
re
Method
The sample for the present research includes 3,382 delinquents committed to state juvenile
na
correctional facilities in Texas between 1987 and 2011. The 3,382 members of this study cohort
represent all offenders sentenced to state juvenile incarceration from 1987-2011 in Texas.2
ur
Unlike regular juvenile court processing and sanctioning in the state, Texas’ blended sentencing
Jo
statute focused on youthful offenders between the ages of 10-16 who have committed one or
more statutorily defined serious and/or violent offenses.3 Upon adjudication under a blended
2
In Texas, there are two general types of offenders who are institutionalized. The first type, indeterminate
commitments, are youth who are sentenced to the Texas Youth Commission (TYC) for a minimum of 9 months and
are broadly more garden variety but still problematic offenders. The second group, and the focus of this research, are
those sentenced to TYC under Texas’s blended sentencing law. This group is referred to as determinate
commitments and may receive maximum sentences ranging from 10-40 years, depending on their offense. Because
of these extended sentence lengths, determinate commitments can be transferred on to the adult system by the time
they reach the maximum age of juvenile correctional jurisdiction. This is not allowed for indeterminate
commitments.
3
During the time frames of this study (1987-2011), the following offenses were eligible for prosecution under
Texas’s blended sentencing statute: murder, attempted murder, capital murder, attempted capital murder, voluntary
manslaughter, intoxication manslaughter, criminally negligent homicide, aggravated kidnapping, attempted
Journal Pre-proof
sentencing prosecution, Texas law allows youthful offenders to receive sentences that can extend
well into adulthood, and which could later result in further incarceration in adult prisons or other
sanctioning in the criminal justice system (e.g., adult probation). Despite the potential for later
adult sanctioning originating from a juvenile-based blended sentence, this research focuses on
the behavioral outcomes of youth during the entire portion of their state juvenile incarceration. In
sum, we examine the misconduct careers of this cohort of offenders during their incarceration in
of
The Texas Youth Commission (TYC) (renamed the Texas Juvenile Justice Department) provided
ro
de-identified data on youthful offenders across a variety of pre-incarceration domains including
-p
but not limited to information on youth demographics, delinquent histories, family-based
re
measures, adverse childhood experiences, and general measures indicative of
antisocial/psychopathic characteristics and other risks. Broadly, these measures derived from a
lP
combination of official records, clinical observations of TYC counselors and correctional staff,
na
and on-site diagnostic examinations that occur at intake to state juvenile correctional facilities.
TYC also provided counts of all forms of officially recorded misconduct during the ward’s state
ur
Measures
violations for the following infractions: attempting, aiding, or abetting a category I (serious) rule
violation, attempting, aiding, or abetting a category II (less serious) rule violation, absconding,
aggravated kidnapping, aggravated sexual assault, attempted aggravated sexual assault, sexual assault, attempted
sexual assault, aggravated assault, aggravated robbery, attempted aggravated robbery, felony injury to a child or
elderly or disabled person, felony deadly conduct, aggravated or first degree controlled substance felony, criminal
solicitation/conspiracy of a capital or first degree felony, second degree felony indecency with a child, criminal
solicitation of a minor, first degree felony arson, habitual felony conduct.
Journal Pre-proof
disruption of program, dress code violation, refusing a drug screen, escape, extortion, fleeing
apprehension, throwing bodily fluids, gambling, hostage taking, indecent exposure, lending,
lying, missed scheduled activity, participation in a riot, inappropriate sexual contact, stealing
over $50, stealing under $50, tampering with security equipment, tattooing, threaten to harm self,
vandalism, violate any law, and violation of security. Total misconduct was dichotomized at the
99th percentile (No = 0, Yes = 1) to capture the most prolific misconduct perpetrators. The top
of
1% accounted for 11% of all misconduct.
ro
Violent misconduct (M = 3.89, SD = 6.49, range = 0-82) included: assaults on staff, assaults on
-p
staff with bodily injury, assault on staff/offender contact, assault by threat of imminent bodily
re
injury, assault of a student, assault of a student with bodily injury, assault of a student/offender
contact, a fight with bodily injury, a fight with no injury, and threatening another with weapon.
lP
Violent misconduct was dichotomized at the 99th percentile (No = 0, Yes, = 1) to capture the
na
most prolific violence perpetrators. The top 1% accounted for 51% of all violent misconduct.
Gang activity. Two measures of gang activity were included based on the significant linkages
ur
between both gang activity and prison violence (Cihan, Reidy, Sorensen, & Chism, 2020;
Jo
Cunningham & Sorensen, 2007; Fahmy, Jackson, Pyrooz, & Decker, 2020; Griffin & Hepburn,
2006; Redi, 2017) and institutional misconduct (DeLisi, Spruill, Peters, Caudill, & Trulson,
2013; Drury & DeLisi, 2011; Trulson, 2007). These were a dichotomous term indicating whether
the youth’s commitment offense was gang-related and a ratio-level indicator of gang activity
Delinquent career. Copious research indicates that youth with more extensive, severe, and earlier
emerging antisocial conduct and juvenile justice system involvement engage in more frequent
Journal Pre-proof
and severe misconduct (Blackburn & Trulson, 2010; Craig & Trulson, 2019; Rembert,
Prior out of home placements is a summary measure of times that the youth was removed from
the home prior to TYC commitment. Total court referrals is a summary measure of referrals to
the juvenile court. Truancy history is a dichotomous term indicating whether the youth had a
of
Adverse childhood experiences. Emerging research (Craig & Trulson, 2019; DeLisi et al., 2010;
ro
Klatt & Kliem, 2019; Rembert, Threadcraft-Walker, Henderson, & Jackson, 2018) indicates that
-p
youth with greater trauma histories characterized by assorted incidents of abuse and neglect
re
experience more misconduct during confinement. Dichotomous terms indicated whether the
youth experienced emotional abuse, abandonment, medical neglect, supervision neglect, sexual
lP
Psychopathology and risk. Externalizing psychopathology (Abbiati, Palix, Gasser, & Moulin,
2019; Choi & Dulisse, 2020; Gardner et al., 2015; Skopp, Edens, & Ruiz, 2007; Thomson et al.,
ur
2019), suicidal thoughts and behaviors (DeLisi et al., 2010), and substance use problems
Jo
(Houser, Belenko, & Brennan, 2012; Jiang, 2005) have been shown to be significantly associated
with institutional misconduct and violence. Five dichotomous measures indicating whether the
youth was a danger to self, danger to others, violent toward their family, suicidal, and had a
substance abuse problem are included in the data. In addition, an ordinal variable of whether the
youth was reared in a chaotic home (no = 0, 1 = somewhat, 2 = very much) was also included.
Butler, & Ellison, 2014; Valentine et al., 2015), sex (Gover, Pérez, & Jennings, 2008; Steiner et
Journal Pre-proof
al., 2014; Steiner & Wooldredge, 2014), race (Berg & DeLisi, 2006; Bonner, Rodrigurez, &
Sorensen, 2017; Steiner & Wooldredge, 2009), and educational attainment (Engstrom & Scott,
2020; Morris et al., 2012) are associated with institutional misconduct and violence.
Dichotomous terms for male, white, black, and continuous terms for age at TYC commitment
Conviction offense. There is mixed evidence for the effects of conviction offense on misconduct
and prison violence (Butler, 2019; Drury & DeLisi, 2011; Steiner et al., 2014), but it is an
of
important consideration. Dichotomous terms for aggravated robbery, aggravated sexual assault,
ro
and murder were included.
-p
Additional covariates that have previously been linked to institutional misconduct and violence
re
include sentence length, time served (years served in custody of the Texas Youth Commission),
chemical dependency treatment, sex offender treatment, mental health treatment, serious and
na
violent offender treatment (Kuo & Zhao, 2010; Steiner et al., 2014; Trulson et al., 2016).
Analytical strategy
To develop a profile of the most disruptive and violent wards, we employed an iterative
analytical strategy. First, binary logistic regression models with odds ratios specified were
executed for total misconduct and violent misconduct. Using estat classification and lroc
commands in Stata 12.1, we produced estimates of the classification accuracy of the model in
terms of cases correctly identified and ROC-AUC models (shown in the figures). Second, we
executed another binary logistic regression model for each dependent variable using only the
Journal Pre-proof
significant covariates from the full model and again assessed the classification accuracy of the
model. This allowed us to assess whether the reduced model provided a more parsimonious
model of disruptive/violent inmates that still maintained good classification accuracy. Third, we
performed sensitivity analyses using the 90th percentile threshold for total and violent
Findings
Logistic Regression Model for 99th Percentile for Total Misconduct (Full Model)
of
Table 2 contains output from the logistic regression model for the 99th percentile for total
ro
misconduct.4 Total adjudications (OR = 1.31, SE = .15, z = 2.46) was significantly associated
-p
with being in the 99th percentile indicating that each additional adjudication in the ward’s
re
delinquent career was associated with 31% increased odds of being chronically noncompliant in
prison. Three demographic factors were significant. African Americans had 812% increased
lP
odds of 99th percentile misconduct (OR = 9.12, SE = 7.57, z = 2.67). Females had 79%
na
increased odds of 99th percentile misconduct (OR = .21, SE = .17, z = -1.97) and age was
inversely associated. With each year increase in age came a reduction in the odds of being in the
ur
99th percentile misconduct group by 45% (OR = .55, SE = .12, z = -2.66). Gang activity during
Jo
confinement was conducive to high levels of total misconduct with each incident conferring
135% increased odds (OR = 2.35, SE = .29, z = 6.99) of 99th percentile status. Clients who
participated in multiple forms of treatment (OR = 2.03, SE = .45, z = 3.18), who were dangers to
self (OR = 3.13, SE = 1.53, z = 2.33), or whose instant offense was aggravated sexual assault
(OR = 6.88, SE = 4.79, z = 2.78) were positively associated with 99th percentile total
misconduct. Although there were several covariates with null associations, the model correctly
4
Instant (current) offense for murder predicted failure perfectly and dropped from the model that resulted in 344 lost
observations. As a result, that covariate is not included in the model.
Journal Pre-proof
classified 99.02% of cases and had excellent classification accuracy (AUC = .97) as shown in the
To assess the sensitivity of the model and examine if a reduced set of covariates produced
changes in classification accuracy, another logistic regression model was executed that included
only the eight significant covariates. As shown in Table 3, all covariates remained significant and
in most cases the effect sizes were substantively similar. However, the inverse age effect
of
intensified, the effect size for African Americans reduced substantially (∆OR from 9.12 to 2.58),
ro
and the effect size for instant offense aggravated sexual assault increased substantially (∆OR
-p
from 6.88 to 8.27). Moreover, the classification accuracy of the model declined only slightly to
re
98.98% cases correctly classified and an AUC = .96 (see Figure 2). In short, the reduced model
was much more efficient than the full model but had nearly the same accuracy.
lP
Table 4 presents output for the logistic regression model of the 99th percentile for violent
Jo
misconduct. Two adverse childhood experiences were significant albeit in different directions.
Emotional abuse was negatively (OR = .63, SE = .14, z = -2.14) and abandonment positively
(OR = 1.55, SE = .32, z = 2.12) associated with violent misconduct at the 99th percentile. Males
had 138% higher odds (OR = 2.38, SE = .83, z = 2.50), African Americans had 70% higher odds
(OR = 1.70, SE = .25, z = 3.60), and age had a strong inverse association (OR = .52, SE = .04, z
= -9.43) with 99th percentile violent misconduct. Each incident of gang activity during
confinement (OR = 2.24, SE = .16, z = 10.96) conferred 114% higher odds of frequent violent
Journal Pre-proof
misconduct. Wards who participated in multiple forms of treatment (OR = 1.47, SE = .10, z =
5.54) or who were dangers to self (OR = 1.72, SE = .26, z = 3.53) were positively associated
with 99th percentile violent misconduct. Those offenders whose commitment offense was murder
were less likely (OR = .49, SE = .15, z = -2.35) to be classified in the 99th percentile violent
misconduct group. Two additional covariates trended toward significance at p<.06 with inverse
associations between years served in TYC custody and sentence length and more frequent
violence. The model correctly classified 90.27% of cases and had strong classification accuracy
of
(AUC = .82) as shown in the ROC curve (Figure 3).
ro
***Figure 3 about here***
-p
To assess the sensitivity of the model, we specified another logistic regression model with only
re
the nine significant covariates. As shown in Table 5, all covariates remained significant with the
exception of emotional abuse that fell from significance. The effect sizes for all variables were
lP
substantively similar with the exception of the commitment offense of murder, which decreased
na
from 0.49 to 0.31 odds. The classification accuracy of the model declined only slightly to
90.15% of cases correctly classified and an AUC = .80 (Figure 3). In short, the reduced model
ur
was much more efficient than the full model but had nearly the same accuracy.
Jo
Sensitivity analyses using the 90th percentile threshold for total and violent misconduct
Logistic regression models for total and violent misconduct at the 90th percentile provided a
sensitivity check for the models set at the 99th percentile. Several of the significant effects
remained; however, there were also some differences. African Americans, multiple forms of
treatment, gang activity during confinement, danger to self, and age remained significant at the
90th percentile for total misconduct. Physical abuse (OR = 1.52, SE = .32, z = 2.00), prior
Journal Pre-proof
placements (OR = 1.20, SE = .07, z = 3.03), and sentence length (OR = .99, SE = .004, z = -3.07)
were significant at the new threshold and instant offense for aggravated sexual assault, total
adjudications, and female dropped from significance. Classification accuracy for the model at the
90th percentile was lower with 91.72% of cases correctly classified and AUC = .86. In the
logistic regression model for violent misconduct at the 90th percentile, the results were mostly the
same. Emotional abuse, abandonment, male, age, African American, danger to self, sentence
length, gang activity during confinement, and multiple forms of treatment were significant at
of
both specifications. The only new effect was that instant offense for murder fell from
ro
significance. Overall, the model had similar classification accuracy with 91.23% of cases
The research goal was to assess the relevant factors in identifying the most frequent and violent
lP
at levels in the 99th percentile within a cohort of serious state institutionalized delinquents. In
estimating full and then more parsimonious logistic regression models, the analyses were able to
ur
ferret out some consistent factors across both models in terms of the most frequent and violent
Jo
misconduct perpetrators in this cohort of serious state delinquents. The findings indicate that
younger age, African-American status, gang activity during confinement, being a danger to self,
and involvement in multiple forms of treatment evince a higher risk of engaging in the 99th and
First, consistent with criminal career and developmental psychopathology research (e.g.,
Moffitt, 1993; Moffitt & Caspi, 2001; Vaughn et al., 2011; Vaughn et al., 2014; Wolfgang,
Figlio, & Sellin, 1972), there is clear evidence that the most severe offenders account for a
Journal Pre-proof
the 90th percentile account for 49% of total misconduct and 50% of violent misconduct and youth
in the 99th percentile account for 11% of total misconduct and 51% of violent misconduct.
Moreover, the top 1% has its most detrimental impact in terms of violence relative to its
with prior studies of both adult prisoners (DeLisi, 2003) and juveniles (Trulson et al., 2010).
Second, although there was consistency in the models regarding the stability of effects at
of
both the 90th and 99th percentiles for several variables, that was not the case for others especially
ro
sex. Females were more likely to be in the 99th percentile for total misconduct, but no effect was
-p
found at the 90th percentile whereas males were more likely to be violent in all models. What
explains these discrepant findings? In the 99th percentile specification, two females had very high
re
misconduct—407 and 642 violations respectively, but aside from these, all of the remaining most
lP
chronic misconduct violators were males. Using the 90th percentile specification removed the
na
influence of these outliers. However, there are also substantive reasons to expect sex differences
in extreme misconduct. There is ample evidence that female correctional clients have worse and
ur
more extensive victimization histories and greater psychopathology (DeLisi et al., 2010; Kuo &
Jo
Zhao, 2019; Warren & Burnette, 2012; Warren et al., 2002; Warren, Hurt, Loper, & Chauhan,
2004; Warren et al., 2004)—conditions that complicate adjustment to the rigors of institutional
life. Moreover, trauma history and some of its downstream sequelae including personality
disorders contribute to an interpersonal style that is susceptible to conflict and other aversive
exchange in correctional environments (Loper, Mahmoodzadegan, & Warren, 2008; Warren &
Burnette, 2012). This is likely particularly true in these data since the wards are juvenile
delinquents and adverse childhood experiences are more proximate in time relative to adult
Journal Pre-proof
female prisoners. The current findings are consonant with prior research. For example, in a
comparative study of female and male prisoners, McCorkle (1995) reported evidence of a
“disturbed and disruptive” prisoner adaption among women but not men in part due to
Third, youth who were dangerous to self were significantly more likely to be in the 99th (and
90th) percentiles for both total misconduct and violence. This comports with Slade’s (2018;
Slade, Forrester, & Baguley, 2020) concept of dual harm conduct characterized as prisoners who
of
engage in both externalized and internalized violence, that is, they hurt themselves and other
ro
inmates. Related, youth who received multiple forms of treatment were significant for both total
-p
and violent misconduct across model specifications, which speaks to the fact that youth with
re
multiple comorbidities (e.g., substance abuse issues and mental health issues) are those who tend
to display lesser ability to adapt to the institutional routine (Houser, Belenko, & Brennan, 2012;
lP
Fourth, consistent with both the adult and juvenile misconduct literature (Steiner et al., 2014;
Valentine et al., 2015), we found consistent effects for the impact of young age on misconduct
ur
perpetration irrespective of the percentile. From a practical standpoint, attention to the youngest
Jo
of all wards may bring dividends with respect to reducing frequent misconduct perpetration.
While the data did not allow us to uncover more specific dynamics for misconduct perpetration
sophisticated wards or if the younger wards simply have a more difficult time adapting to the
institutional routine, or both—this seems a fruitful line of policy relevant research. This finding
is also relevant with regard to gang activity and its relation to misconduct. Several dynamics
might explain why gang members are more involved in misconduct, such as gang disputes,
Journal Pre-proof
greater attention by correctional officers, and peer pressure from older and more sophisticated
gang members. Although we did not have information with respect to the prevalence of these and
other potential dynamics, this type of information would shed additional light on their
misconduct careers behind bars. Moreover, the consistent effects of African-American race on
misconduct are worthy of further attention. Steiner and Wooldredge (2009) observed that
individuals from disadvantaged communities may have problems adjusting to the rigid structure
of prison life, and that attitudes toward violence in the community also influence attitudes toward
of
violence in prisons (also see, Choi & Dulisse, 2020). A research opportunity is to examine
ro
whether high levels of total and violent misconduct are the result of greater rule enforcement,
-p
greater perpetration, or some explanation in between. Such a research implication might also pay
re
specific attention to the types of misconduct perpetrated, especially violent misconduct, which
We recognize limitations of the current study. First, this research focused on an extremely
na
serious and mostly violent cohort of institutionalized juvenile offenders. Because Texas’ blended
sentencing law is limited to a defined set of serious and/or violent offenders, the blended
ur
sentenced offender sample materializes from a form of selection bias at the legal and prosecution
Jo
stage in that these offenders are not garden-variety state delinquent commitments. This impacts
the generalizability of our study albeit that this group does not generalize to other delinquents is
less of a concern as these are the types of offenders of most consequence for juvenile
for involvement in institutional misconduct, additional variables and specificity among variables
would have perhaps improved our ability to explain why some factors were important to
membership in the 99th and 90th percentiles. For example, regarding significant effects for age, it
Journal Pre-proof
would have been interesting to see if misconduct relates to a potential victimization situation,
gang activity during confinement, it would have been useful to understand if misconduct
perpetration, for example, was a consequence of pressures by some gang members on others
(e.g., pressure from older gang members on young gang members), or geographical pressure
(e.g., youth from Houston competing with youth from Dallas) to perpetrate misconduct. These
are realities in juvenile institutional environments that warrant empirical attention. Third, as with
of
any official data source, especially behavioral counts of individuals under confinement, we were
ro
not able to address factors such as the detection of misconduct incidents or differences in
-p
correctional officer reporting at different institutions, among other relevant considerations in
re
accounting for misconduct (Hewitt, Poole & Regoli, 1984). This limitation is less central to
violent misconduct perpetration than total misconduct perpetration as more serious incidents
lP
(e.g., assault on a ward or an officer) are more likely to be discovered and reported, but this is a
na
limitation that if able to be accounted for would have improved this study.
Conclusion
ur
Absent these and other potential limitations to this research, we believe this research fills an
Jo
important gap with respect to involvement in institutional misconduct among serious and/or
violent delinquent offenders. By focusing on the extremes of misconduct perpetrators, our goal
was to dig deeper into the misconduct careers of some of the most problematic institutionalized
delinquents so to uncover the factors related to these more extreme levels of misconduct
perpetration “within” a group of already serious and often violent offenders. The overarching
finding of this study is that delinquent careers do not stop at the gates of juvenile correctional
References
Abbiati, M., Palix, J., Gasser, J., & Moulin, V. (2019). Predicting physically violent misconduct
in prison: A comparison of four risk assessment instruments. Behavioral Sciences & the
Associated Press. (2020). California inmate writes letter to newspaper, says he killed 2 molesters
of
in prison. Retrieved April 13, 2020, from
ro
https://www.usatoday.com/story/news/nation/2020/02/22/inmate-jonathan-watson-says-he-
killed-2-molesters-california-prison/4842042002/
-p
re
Berg, M. T., & DeLisi, M. (2006). The correctional melting pot: Race, ethnicity, citizenship, and
Blackburn, A. G., & Trulson, C. R. (2010). Sugar and spice and everything nice? Exploring
na
institutional misconduct among serious and violent female delinquents. Journal of Criminal
Bonner, H. S., Rodriguez, F. A., & Sorensen, J. R. (2017). Race, ethnicity, and prison
Jo
Butler, H. D., Griffin III, O. H., & Johnson, W. W. (2013). What makes you the “worst of the
Butler, H. D., Johnson, W. W., & Griffin III, O. H. (2014). The treatment of the mentally ill in
Journal Pre-proof
supermax facilities: An evaluation of state supermax policies. Criminal Justice and Behavior,
41(11), 1338-1353.
Carson, E. A., & Cowhig, M. P. (2020). Mortality in state and federal prisons, 2001-2016—
Statistical tables. Washington, DC: U.S. Department of Justice, Office of Justice Programs,
Choi, J., & Dulisse, B. (2020). The importation of violent “codes” of South Korean inmates. The
of
Cihan, A., Davidson, M., & Sorensen, J. (2017). Analyzing the heterogeneous nature of inmate
ro
behavior: Trajectories of prison misconduct. The Prison Journal, 97(4), 431-450.
-p
Cihan, A., Reidy, T. J., Sorensen, J., & Chism, K. A. (2020). Assessing the developmental
re
patterns of visitation on prison misconduct: do visitation patterns matter? Criminal Justice
Cihan, A., & Sorensen, J. R. (2019). Examining developmental patterns of prison misconduct:
na
Cihan, A., Sorensen, J., & Chism, K. A. (2017). Analyzing the offending activity of inmates:
Jo
50, 12-18.
Cochran, J. C., & Mears, D. P. (2017). The path of least desistance: Inmate compliance and
Cochran, J. C., Mears, J. P., Bales, W. D., & Stewart, E. A. (2014). Does inmate behavior affect
Craig, J. M., & Trulson, C. R. (2019). Continuity of the delinquent career behind bars: Predictors
of violent misconduct among female delinquents. Aggression and Violent Behavior, 49, 101301.
Cunningham, M. D., & Sorensen, J. R. (2007). Predictive factors for violent misconduct in close
DeLisi, M. (2003). Criminal careers behind bars. Behavioral Sciences & the Law, 21(5), 653-
of
669.
ro
DeLisi, M., Berg, M. T., & Hochstetler, A. (2004). Gang members, career criminals and prison
-p
violence: Further specification of the importation model of inmate behavior. Criminal Justice
re
Studies, 17(4), 369-383.
DeLisi, M., Drury, A. J., Kosloski, A. E., Caudill, J. W., Conis, P. J., Anderson, C. A., ... &
lP
Beaver, K. M. (2010). The cycle of violence behind bars: Traumatization and institutional
na
misconduct among juvenile delinquents in confinement. Youth Violence and Juvenile Justice,
8(2), 107-121.
ur
DeLisi, M., Spruill, J. O., Peters, D. J., Caudill, J. W., & Trulson, C. R. (2013). “Half in, half
Jo
out:” Gang families, gang affiliation, and gang misconduct. American Journal of Criminal
DeLisi, M., Trulson, C. R., Marquart, J. W., Drury, A. J., & Kosloski, A. E. (2011). Inside the
prison black box: Toward a life course importation model of inmate behavior. International
DeLisi, M., Vaughn, M. G., Beaver, K. M., Wright, J. P., Hochstetler, A., Kosloski, A. E., &
Drury, A. J. (2008). Juvenile sex offenders and institutional misconduct: The role of thought
Journal Pre-proof
Drury, A. J., & DeLisi, M. (2011). Gangkill: An exploratory empirical assessment of gang
membership, homicide offending, and prison misconduct. Crime & Delinquency, 57(1), 130-146.
Engstrom, R. S., & Scott, D. (2020). Juvenile institutional misconduct: Examining the role of
10.1177/0011128719901113.
Fahmy, C., Jackson, D. B., Pyrooz, D. C., & Decker, S. H. (2020). Head injury in prison: Gang
of
membership and the role of prison violence. Journal of Criminal Justice, 101658.
ro
Fox, V. (1958). Analysis of prison disciplinary problems. Journal of Criminal Law, Criminology
Gover, A. R., Pérez, D. M., & Jennings, W. G. (2008). Gender differences in factors contributing
Griffin, M. L., & Hepburn, J. R. (2006). The effect of gang affiliation on violent misconduct
Jo
among inmates during the early years of confinement. Criminal Justice and Behavior, 33(4),
419-466.
Hewitt, J. D., Poole, E. D., & Regoli, R. M. (1984). Self-reported and observed rule-breaking in
Houser, K. A., Belenko, S., & Brennan, P. K. (2012). The effects of mental health and substance
abuse disorders on institutional misconduct among female inmates. Justice Quarterly, 29(6),
799-828.
Journal Pre-proof
Irwin, J., & Cressey, D. R. (1962). Thieves, convicts and the inmate culture. Social Problems,
10(2), 142-155.
Jiang, S. (2005). Impact of drug use on inmate misconduct: A multilevel analysis. Journal of
Klatt, T., & Kliem, S. (2019). The influence of harsh parenting and parental warmth during
doi: 10.1177/0886260518820675.
of
Kuanliang, A., Sorensen, J. R., & Cunningham, M. D. (2008). Juvenile inmates in an adult prison
ro
system: Rates of disciplinary misconduct and violence. Criminal Justice and Behavior, 35(9),
1186-1201.
-p
re
Kuo, S. Y., & Zhao, R. (2019). Prison misconduct among female inmates with mental disorders,
Lanes, E. C. (2011). Are the “worst of the worst” self-injurious prisoners more likely to end up
Loper, A. B., Mahmoodzadegan, N., & Warren, J. I. (2008). Childhood maltreatment and cluster
Lovell, D. (2008). Patterns of disturbed behavior in a supermax population. Criminal Justice and
MacKenzie, D. L. (1987). Age and adjustment to prison: Interactions with attitudes and anxiety.
Marquart, J. W., & Crouch, B. M. (1984). Co-opting the kept: Using inmates for social control in
Journal Pre-proof
male and female mentally ill prison inmates. Journal of Criminal Justice, 23(1), 53-61.
Moffitt, T. E., & Caspi, A. (2001). Childhood predictors differentiate life-course persistent and
of
Psychopathology, 13(2), 355-375.
ro
Morris, R. G., Carriaga, M. L., Diamond, B., Piquero, N. L., & Piquero, A. R. (2012). Does
-p
prison strain lead to prison misbehavior? An application of general strain theory to inmate
re
misconduct. Journal of Criminal Justice, 40(3), 194-201.
Myers, L. B., & Levy, G. W. (1978). Description and prediction of the intractable inmate.
lP
Petersilia, J., Honig, P., & Hubay, C. (1980). The prison experience of career criminals. Santa
Poole, E. D., & Regoli, R. M. (1979). Race, institutional rule breaking, and disciplinary
Jo
response: A study of discretionary decision making in prison. Law & Society Review, 14, 931-
941.
Reidy, T. J., Cihan, A., & Sorensen, J. R. (2017). Women in prison: Investigating trajectories of
Reidy, T. J., Sorensen, J. R., & Cihan, A. (2018). Institutional misconduct among juvenile
Journal Pre-proof
Rembert, D. A., Threadcraft-Walker, W., Henderson, H., & Jackson, R. (2018). Predicting youth
assault and institutional danger in juvenile correctional facilities. Journal of Criminal Justice,
58(C), 47-55.
Skopp, N. A., Edens, J. F., & Ruiz, M. A. (2007). Risk factors for institutional misconduct
of
among incarcerated women: An examination of the criterion-related validity of the Personality
ro
Assessment Inventory. Journal of Personality Assessment, 88(1), 106-117.
-p
Slade, K. (2018). Dual harm: an exploration of the presence and characteristics for dual violence
re
and self-harm behaviour in prison. Journal of Criminal Psychology, 8(2), 97-111.
Slade, K., Forrester, A., & Baguley, T. (2020). Co-existing violence and self-harm: Dual harm in
lP
Steiner, B., Butler, H. D., & Ellison, J. M. (2014). Causes and correlates of prison inmate
misconduct: A systematic review of the evidence. Journal of Criminal Justice, 42(6), 462-470.
ur
Steiner, B., & Wooldredge, J. (2009). The relevance of inmate race/ethnicity versus population
Jo
composition for understanding prison rule violations. Punishment & Society, 11(4), 459-489.
Steiner, B., & Wooldredge, J. (2014). Sex differences in the predictors of prisoner misconduct.
Thomson, N. D., Towl, G. J., & Centifanti, L. (2016). The habitual female offender inside: How
psychopathic traits predict chronic prison violence. Law and Human Behavior, 40(3), 257-269.
Thomson, N. D., Vassileva, J., Kiehl, K. A., Reidy, D., Aboutanos, M., McDougle, R., & DeLisi,
M. (2019). Which features of psychopathy and impulsivity matter most for prison violence? New
Journal Pre-proof
evidence among female prisoners. International Journal of Law and Psychiatry, 64, 26-33.
Toch, H., Adams, K., & Greene, R. (1987). Ethnicity, disruptiveness, and emotional disorder
Trulson, C. R., DeLisi, M., Caudill, J. W., Belshaw, S., & Marquart, J. W. (2010). Delinquent
of
Trulson, C.R., DeLisi, M. & Marquart, J.W. (2011). Institutional misconduct, delinquent
ro
background, and re-arrest frequency among serious and violent delinquent offenders. Crime &
sentencing, second chances, and the Texas youth commission. Austin, TX: University of Texas
lP
Press.
na
Valentine, C. L., Mears, D. P., & Bales, W. D. (2015). Unpacking the relationship between age
Vaughn, M. G., DeLisi, M., Gunter, T., Fu, Q., Beaver, K. M., Perron, B. E., & Howard, M. O.
Jo
(2011). The severe 5%: A latent class analysis of the externalizing behavior spectrum in
Vaughn, M. G., Salas-Wright, C. P., DeLisi, M., & Maynard, B. R. (2014). Violence and
externalizing behavior among youth in the United States: Is there a severe 5%? Youth Violence
Warren, J. I., & Burnette, M. (2012). Factor invariance of Cluster B psychopathology among
male and female inmates and association with impulsive and violent behavior. The Journal of
Journal Pre-proof
Warren, J. I., Burnette, M., South, S. C., Chauhan, P., Bale, R., & Friend, R. (2002). Personality
disorders and violence among female prison inmates. Journal of the American Academy of
Warren, J. I., Hurt, S., Loper, A. B., & Chauhan, P. (2004). Exploring prison adjustment among
female inmates: Issues of measurement and prediction. Criminal Justice and Behavior, 31(5),
624-645.
of
Wolfgang, M. E. (1961). Quantitative analysis of adjustment to the prison community. Journal of
ro
Criminal Law, Criminology, & Police Science, 51(6), 607-618.
-p
Wolfgang, M. E., Figlio, R. M., & Sellin, T. (1972). Delinquency in a birth cohort. Chicago, IL:
re
University of Chicago Press.
Wood Jr., B. S., Wilson, G. G., Jessor, R., & Bogan, J. B. (1966). Troublemaking behavior in a
lP
Zink, T. M. (1957). Are prison troublemakers different? The Journal of Criminal Law,
ur
of
Instant Offense Aggravated Sexual Assault 21% 79%
Instant Offense Murder 10% 90%
ro
Male 94% 6%
Black 39% 61%
White 20% 80%
Gang-Related Offense
Truancy History -p 17% 83%
71% 29%
re
Emotional Abuse 19% 81%
Abandonment 11% 89%
Medical Neglect 4% 96%
lP
Table 2. Logistic Regression Model for 99th Percentile for Total Misconduct (Full Model)
Variable OR SE z
Total Adjudications 1.31 .15 2.46*
Prior Out of Home Placements 1.21 .24 0.95
Total Court Referrals .80 .16 -1.11
Emotional Abuse .78 .55 -0.35
Abandonment .29 .23 -1.54
Medical Neglect .99 1.03 -0.01
Supervision Neglect .49 .33 -1.07
Sexual Abuse 1.98 1.18 1.14
Physical Neglect 1.91 1.49 0.83
Physical Abuse 1.04 .70 0.07
Highest Grade Completed .82 .16 -1.00
of
Youth Has Substance Abuse .53 .28 -1.19
Problem
ro
Truancy History .44 .22 -1.62
Male .21 .17 -1.97*
Age .55 .12 -2.66**
Years Served in TYC
African American -p
1.13
9.12
.20
7.57
0.67
2.67**
re
White 4.03 3.75 1.50
Sentence Length .98 .02 -1.18
Gang-Related Offense .60 .62 -0.50
lP
Assault
Youth is Suicidal 1.58 .99 0.72
2
Model χ 200.54***
Pseudo R2 .515
Correctly Classified 99.02%
*p<.05, **p<.01, ***p<.001
Journal Pre-proof
Table 3. Logistic Regression Model for 99th Percentile for Total Misconduct (Reduced Model)
Variable OR SE z
Total Adjudications 1.18 .07 2.69**
Male .19 .14 -2.33*
Age .46 .07 -4.95***
African American 2.58 1.09 2.25*
Gang Activity During Confinement 2.30 .25 7.67***
Multiple of Treatment 2.16 .45 3.73***
Danger to Self 3.06 1.26 2.72**
Instant Offense Aggravated Sexual 8.27 3.83 4.57***
Assault
Model χ2 173.10***
2
Pseudo R .444
of
Correctly Classified 98.97%
*p<.05, **p<.01, ***p<.001
ro
Variable
Total Adjudications
OR -p
Table 4. Logistic Regression Model for 99th Percentile for Violent Misconduct (Full Model)
1.05
SE
.04
z
1.42
re
Prior Out of Home Placements 1.08 .06 1.45
Total Court Referrals .97 .05 -0.54
lP
Problem
Truancy History 1.17 .18 1.00
Male 2.38 .83 2.50**
Age .52 .04 -9.43***
Years Served in TYC .90 .05 -1.88†
African American 1.70 .25 3.60***
White .97 .19 -0.15
Sentence Length .99 .01 -1.88†
Gang-Related Offense .95 .18 -0.28
Gang Activity During Confinement 2.24 .16 10.96***
Multiple forms of Treatment 1.47 .10 5.54***
Danger to Self 1.72 .26 3.53***
Danger to Others .96 .14 -0.25
Chaotic Home 1.17 .11 1.69
Violent Toward Family 1.02 .16 0.14
Journal Pre-proof
of
Table 5. Logistic Regression Model for 99th Percentile for Violent Misconduct (Reduced Model)
Variable OR SE z
Emotional Abuse .79 .14 -1.34
ro
Abandonment 1.64 .31 2.59**
Male 2.25 .75 2.44*
Age
African American
Gang Activity During Confinement
-p .55
1.64
2.39
.03
.20
.17
-11.95***
3.99***
12.30***
re
Multiple Forms of Treatment 1.54 .10 6.38***
Danger to Self 1.90 .27 4.51
lP
Figure 1. Post-Estimation ROC Curve for 99th Percentile Total Misconduct (Full Model)
1.00
0.75
Sensitivity
0.50
of
0.25
ro
0.00
0.00 0.25
na
0.75
ur
Sensitivity
0.50
Jo
0.25
0.00
Figure 3. Post-Estimation ROC Curve for 99th Percentile Violent Misconduct (Full Model)
1.00
0.75
Sensitivity
0.50
of
0.25
ro
0.00
Figure 4 Post-Estimation ROC Curve for 99th Percentile Violent Misconduct (Reduced Model)
1.00
na
0.75
ur
Sensitivity
Jo
0.50
0.25
0.00
Highlights
Younger age, African-American status, gang activity during confinement, being a danger to self,
and involvement in multiple forms of treatment evince a higher risk of engaging in the 99th
percentile of total and violent misconduct regardless of model specification.
Other factors such as commitment offense, adverse childhood experiences, and sex have
differential effects depending on whether the 90th or 99th percentile thresholds are considered.
Youth in the 90th percentile account for 49% of total misconduct and 50% of violent misconduct.
Youth in the 99th percentile account for 11% of total misconduct and 51% of violent misconduct.
of
We discuss implications for policy and practice and the importance of understanding antisocial
development of youthful offenders and their institutional maladjustment.
ro
-p
re
lP
na
ur
Jo